Model providers

Learn about AI model providers on Runware's platform. Understand how we integrate leading models from Google, KlingAI, MiniMax, and other providers through a unified API.

Introduction

Runware integrates with leading AI model providers to offer you access to the best available models for image and video generation. Rather than managing multiple APIs and different authentication systems, our unified platform abstracts these complexities while preserving each provider's unique capabilities.

We support models through two integration approaches: open source models running on our own infrastructure, and closed source models accessed through external provider APIs. This hybrid approach ensures you get optimal performance and pricing while accessing the latest innovations from both the open source community and commercial providers.

How provider integrations work

All models, regardless of provider, use the same core API structure you're familiar with from our platform. Common parameters like width, height, seed, duration, fps, and includeCost work consistently across all providers, making it easy to switch between models or test different options.

When providers offer unique capabilities that don't fit our standard parameters, we expose them through provider-specific settings. These appear in a providerSettings object within your request, organized by provider name.

{
  "taskType": "videoInference",
  "model": "google:3@0",
  "positivePrompt": "a cat playing with yarn",
  "duration": 8,
  "width": 1280,
  "height": 720,
  "providerSettings": {
    "google": {
      "enhancePrompt": true,
      "generateAudio": true
    }
  }
}

This design preserves the simplicity of our unified API while giving you access to provider-specific features when you need them. You can use the same basic parameters for quick testing, then add provider settings for advanced control.

Provider abstractions and limitations

While our unified API normalizes most parameters across providers, each integration has specific capabilities and constraints. These may include:

  • Supported generation modes: Not all providers support every workflow (text-to-image, image-to-video, etc).
  • Parameter ranges: Image aspect ratios, video resolution and durations options, and other technical specifications vary.
  • Input requirements: While formats are standardized, providers may impose different limits on resolution, file size, or other requirements depending on the model or workflow.
  • Feature availability: Advanced capabilities like audio generation, motion controls, or style effects are provider-specific.

When switching between providers, always check the provider-specific documentation to ensure your parameters and workflows are supported.

Getting started

Each provider page includes detailed documentation of their specific capabilities, parameter options, and usage examples. Start with the provider that best matches your immediate needs, then explore others as your requirements evolve.

Available provider documentation:

For general API usage and authentication, see our introduction page. For workflow-specific details, visit the image inference and video inference introduction pages.